Factors That Limit Positron Emission Tomography Imaging of P

Apr 18, 2013 - barrier, such as P-glycoprotein (P-gp) and breast cancer resistance protein ..... A three-dimensional reconstruction of brain capillari...
0 downloads 0 Views 2MB Size
Review pubs.acs.org/molecularpharmaceutics

Terms of Use

Factors That Limit Positron Emission Tomography Imaging of P‑Glycoprotein Density at the Blood−Brain Barrier Pavitra Kannan,†,‡ Victor W. Pike,† Christer Halldin,‡ Oliver Langer,§,∥ Michael M. Gottesman,⊥ Robert B. Innis,*,† and Matthew D. Hall⊥ †

Molecular Imaging Branch, National Institute of Mental Health, National Institutes of Health, Bethesda, Maryland 20892, United States ‡ Department of Clinical Neuroscience, Karolinska Institutet, 17176 Stockholm, Sweden § Health and Environment Department, Biomedical Systems, AIT Austrian Institute of Technology GmbH, 2444 Seibersdorf, Austria ∥ Department of Clinical Pharmacology, Medical University of Vienna, 1090 Vienna, Austria ⊥ Laboratory of Cell Biology, National Cancer Institute, National Institutes of Health, Bethesda, Maryland 20892, United States ABSTRACT: Efflux transporters located at the blood−brain barrier, such as P-glycoprotein (P-gp) and breast cancer resistance protein (BCRP), regulate the passage of many drugs in and out of the brain. Changes in the function and density of these proteins, in particular P-gp, may play a role in several neurological disorders. Several radioligands have been developed for measuring P-gp function at the blood−brain barrier of human subjects with positron emission tomography (PET). However, attempts to measure P-gp density with radiolabeled inhibitors that bind to these proteins in vivo have not thus far provided useful, quantifiable PET signals. Herein, we argue that not only the low density of transporters in the brain as a whole but also their very high density in brain capillaries act to lower the concentration of ligand in the plasma and thereby contribute to absent or low signals in PET studies of P-gp density. Our calculations, based on published data and theoretical approximations, estimate that whole brain densities of many efflux transporters at the blood−brain barrier range from 0.04 to 5.19 nM. We conclude that the moderate affinities (>5 nM) of currently labeled inhibitors may not allow measurement of efflux transporter density at the blood−brain barrier, and inhibitors with substantially higher affinity will be needed for density imaging of P-gp and other blood−brain barrier transporters. KEYWORDS: positron emission tomography (PET), ABC transporters, imaging, blood−brain barrier, inhibitors, density



INTRODUCTION

in vivo, quantifying both transporter function and density is a significant and important challenge. Molecular imaging techniques such as positron emission tomography (PET) offer the potential for in vivo measurement of function and density of protein targets. In PET, a radiotracer is injected at a subpharmacological dose, and biomathematical modeling is applied to acquired data to determine output measures related to the interaction of the radiotracer with the receptor target. In the case of functional studies, increased or decreased uptake of a radiotracer measures protein function. Glucose metabolism in tissue, for example, is reflected by increased uptake of [18F]fluorodeoxyglucose.6 Efflux transport by P-gp or BCRP, on the other hand, is reflected by little to no uptake of the radiolabeled substrate in tissue, and inhibition of efflux transport results in increased substrate accumulation.1 PET

P-Glycoprotein (P-gp, encoded by ABCB1) and breast cancer resistance protein (BCRP, encoded by ABCG2) are among the efflux transporters that are essential to the function of the blood− brain barrier.1 These two transporters are apically located (blood-facing) in the endothelial cells of brain capillaries to prevent the entry of drugs into the brain, thereby protecting it from exposure to toxins in the blood.2,3 As a result, P-gp and BCRP can also impede the entry of potential therapeutics to the brain, and this barrier function may be exacerbated in pathophysiological conditions. Increased function of P-gp, for example, may contribute to drug resistance in epilepsy and may decrease the effectiveness of treating HIV infection of the brain.4,5 Changes in expression (i.e., density) of these transporters may also be clinically relevant because expression affects efflux capacity. For example, loss of P-gp expression would result in net decrease in function and, consequently, in dysfunction of the blood−brain barrier. Because the function and density of Pgp and BCRP in neurological disorders are not well understood © 2013 American Chemical Society

Received: Revised: Accepted: Published: 2222

January 8, 2013 March 27, 2013 April 18, 2013 April 18, 2013 dx.doi.org/10.1021/mp400011g | Mol. Pharmaceutics 2013, 10, 2222−2229

Molecular Pharmaceutics

Review

Table 1. Brain Concentrations of Radiolabeled P-gp and BCRP Inhibitors in Rat and Mouse Measured 30 min after Injection radiotracer uptake in rat brain at 25 min radiolabeled inhibitor 11

[ C]laniquidar [11C]elacridar 1-[18F]fluoroelacridar [18F]fluoroethylelacridar [11C]tariquidar [18F]fluoroethyltariquidar [11C]-1b [11C]-2c

baseline

blocked

unita

0.07

0.6

%ID/g

0.20 0.14

1.08 1.29

SUV SUV

0.19

0.54

SUV

1.2

0.90

SUV

radiotracer uptake in mouse brain at 30 min wild-type

P-gp knockout

BCRP knockout

P-gp/BCRP knockout

unit

0.05 0.18 0.23 0.08 0.02 0.15 0.07

0.15 0.46

0.05 0.20

0.18 0.04 0.31 0.14

0.08 0.02 0.18 0.10

1.0 1.39 1.76 0.70 0.75 1.39 0.60

SUV SUV SUV SUV SUV SUV SUV

0.96

1.17

0.80

1.79

SUV

reference 18 22 16 17 21 20 15 21 19 23

a

%ID/g = % injected dose/gram of tissue. SUV = standardized uptake value, which is the ratio of radioactivity concentration measured in tissue at time t to injected dose at time of injection divided by body weight. A SUV of 1.0 is the value that would be obtained throughout the body for a hypothetical perfectly uniform distribution. b1 = novel P-gp inhibitor, [11C]6,7-dimethoxy-2-{3-[4-methoxy-3,4-dihydro-2H naphthalene-(1E)ylidene]-propyl}-1,2,3,4-tetrahydro-isoquinoline c2 = putative BCRP inhibitor, [11C]methyl 4-((4-(2-(6,7-dimethoxy-1,2,3,4-tetrahydroisoqunolin-2yl)ethyl)phenyl)amino-carbonyl)-2-(quinoline-2-carbonylamino)benzoate

capillary. We propose that the density of P-gp (and BCRP) is high enough within the capillary compartment to substantially affect the free concentration of radiotracer during its 1−2 s transit through the capillaries. Here, we use the results from PET studies of P-gp density to explain how these two factorslow binding potential and high, localized transporter densitymay greatly affect brain signal. In the following sections, we assume for simplicity that: (1) The brain signal measured in PET studies of P-gp density is “binding to P-gp”. In reality, “brain uptake of radioactivity” includes both parenchymal uptake (after passing the blood− brain barrier) and radioactivity in the vascular compartment (about 5% of total brain volume (see below)). PET studies seek to determine the amount of radiolabeled inhibitor bound to P-gp in the brain capillaries. However, PET lacks the resolution to separately measure uptake in parenchyma and that in the vascular compartment. (2) A radiolabeled inhibitor can bind to all the Pgp in brain endothelial cells, but the majority of P-gp is localized at the luminal membrane. While P-gp is constantly recycled on/ off the cell surface, the majority is thought to be expressed at the cell surface.

has been successful in using radiolabeled substrates to measure Pgp function at the blood−brain barrier for at least two reasons. First, each molecule of P-gp can transport multiple substrate molecules. Second, upon inhibition of transport, amplification of the PET signal can occur by trapping some of the substrate in cellular organelles.7,8 In the case of density studies, the density of the target is inferred from the binding potentiala parameter that is the product of the concentration of binding sites (Bmax) and the affinity of the radioligand for the protein (1/KD, where KD is the equilibrium dissociation constant).9 [The term radioligand describes a specific class of compounds that reversibly or irreversibly bind to the protein target (e.g., radiolabeled inhibitor), whereas the term radiotracer refers to a more general class of compounds that includes substrates and radioligands.] Because substrates are transported quickly when in the vicinity of a transporter,10 they cannot be used to measure the density of these efflux transporters by PET. Inhibitors of P-gp or of BCRP, on the other hand, are known to directly bind to their respective transporters10 and might behave akin to antagonist receptor radioligands, thereby measuring density. However, PET has been unsuccessful at measuring P-gp density using radiolabeled inhibitors. Unlike the signal amplification obtained by substrate trapping, the maximal signal that can be obtained by binding is one radiolabeled inhibitor for each transporter (assuming one binding site per transporter). The difficulty in imaging P-gp density is also due to two previously described phenomena. First, many of the inhibitors cross-react with P-gp and BCRP,11 which confounds results obtained in P-gp knockout, Bcrp knockout, or P-gp/Bcrp knockout mice. Second, some inhibitors of P-gp are also substrates for BCRP.12 At very low concentrations, high affinity inhibitors might be transported by P-gp, but evidence for this appears to be conflicting.13,14 The purpose of this perspective is to highlight two other difficulties of imaging P-gp density at the blood−brain barrier using PET. The first is that the density of P-gp in the whole brain has been variably reported in the literature but plays a critical role in the affinity of the radioligand required to have a measurable signal from in vivo PET imaging. This may be exacerbated by the low resolution of PET imaging. The second difficulty, which has not been discussed in the literature, is the high density of P-gp within the local microenvironment (microcompartment) of the



RESULTS FROM PET STUDIES MEASURING P-GP DENSITY PET studies using high-affinity P-gp/BCRP inhibitors to measure transporter density at the blood−brain barrier have yielded confusing results (Table 1). According to conventional PET studies measuring receptor density, the specific binding of a radioligand to its receptor should be displaceable.9 If we assume that a radiolabeled inhibitor binds only to P-gp (i.e., is specific), then high concentrations of unlabeled inhibitor should displace the radioligand bound to P-gp and consequently decrease the measured brain signal. However, brain uptake of all except one of the tested radiolabeled inhibitors was low 30 min after injection in wild-type rats but increased at least 150% after blockade with high doses of nonradiolabeled inhibitors; the increase was not due to altered peripheral metabolism after blockade.15−19 Results obtained from transgenic mice are also not consistent with expectations for a P-gp binding molecule. For example, brain uptake of the “high-affinity” P-gp inhibitor [11C]tariquidar was low in wild-type, P-gp (mdr1a−/−/mdr1b−/−6), and Bcrp(abcg26) knockout mice but was high in P-gp/Bcrp (mdr1a−/−/mdr1b−/−/ 2223

dx.doi.org/10.1021/mp400011g | Mol. Pharmaceutics 2013, 10, 2222−2229

Molecular Pharmaceutics

Review

Table 2. Calculated Values of Receptor Density for Various Efflux Transporters Expressed in Human, Monkey, and Mouse Brainsa human gene

protein

ABC Transporters ABCA2 ABCA8 ABCB1 P-gp ABCC4 Mrp4 ABCG2 Bcrp SLC Transporters SLC1A3 EAAT1 SLC2A1 GLUT1 SLC2A3 GLUT3 SLC3A2 4F2hc SLC6A12 BGT1 SLC7A1 CAT1 SLC7A5 LAT1 SLC16A1 MCT1 SLC19A1 RFC SLC29A1 ENT1 Receptors INSR insulin receptor LRP1 Lrp1 Tf R1 transferrin receptor

monkey calculated Bmax (nM)

protein conc. (fmol/μg)

2.86 1.21 6.06 0.195 4.07

0.61 0.26 1.28 0.04 0.86

NR NR 4.71 0.286 7.10

24.5 139 4.40 3.47 3.16 1.13 0.431 2.27 0.763 0.568

5.19 29.47 0.93 0.74 0.67 0.24 0.09 0.48 0.16 0.12

NR 129 1.22 NR NR NR ULQ 0.834 NR 0.541

1.09 1.51 2.34

0.23 0.32 0.50

1.52 1.29 NR

protein conc. (fmol/μg)

mouse calculated Bmax (nM)

protein conc. (fmol/μg)

calculated Bmax (nM)

1.00 0.06 1.51

ULQ ULQ 14.1 1.59 2.20

2.99 0.34 0.47

27.35 0.26

0.18 0.11 0.32 0.27

NR 90.0 ULQ 16.4 NR NR 2.19 23.7 NR 0.985 1.16 1.07 5.84

19.08 3.48

0.46 5.02 0.21 0.25 0.23 1.24

a

ABCG2/BCRP values are half that reported in Uchida and colleagues,29 to account for the homodimerization of protein product to form a functional unit. ULQ = Under limit quantification. NR = Gene not reported/examined.



abcg26) triple knockout mice.15,20 Tariquidar is a substrate for BCRP,12 which explains its low uptake in wild-type and P-gp knockout mice. However, one would have expected higher uptake in Bcrp knockout mice because P-gp is still expressed in brain capillaries and can bind to [11C]tariquidar. Brain uptake of other radiolabeled inhibitors followed a similar pattern in the four strains of mice.16,17,21,22 Although extensive work has been done on P-gp inhibitors, only one putative BCRP inhibitor has been radiolabeled and tested in animals. The brain uptake of this 11C-labeled compound (chemical name [11C]methyl 4-((4-(2-(6,7-dimethoxy-1,2,3,4tetrahydroisoquinolin-2-yl)ethyl)phenyl)amino-carbonyl)-2(quinoline-2-carbonylamino)benzoate23) in wild-type mice was high but was even higher in P-gp and P-gp/BCRP knockout mice, similar to results obtained with most of the P-gp inhibitors.23 Among these eight reported ligands, only one inhibitor showed a different uptake profile. The P-gp inhibitor [11C]-1 (chemical name 6,7-dimethoxy-2-{3-[4-methoxy-3,4-dihydro2H naphthalene-(1E)-ylidene]-propyl}-1,2,3,4-tetrahydro-isoquinoline19) had a high brain uptake (>1.0 SUV) in rats at baseline conditions and a 30% reduced uptake after blockade of transporters.19 While these results are what one might expect from a PET study measuring efflux transporter density, they should be interpreted cautiously. The inhibitor [11C]-1 is approximately 2 orders of magnitude less potent than tariquidar and elacridar24 and may not, therefore, have the affinity required to specifically measure the density of P-gp at the blood−brain barrier (see discussion below).

TWO FACTORS AFFECT BRAIN SIGNAL IN PET STUDIES OF P-GP DENSITY Low Binding Potential of Efflux Transporters. Even after blockade or genetic knockout of transporters, the raw values of brain uptake for most of the radiolabeled inhibitors are still lower (5 for binding potential (Bmax/KD) is generally recommended to achieve measurable brain uptake; however, the field lacks consensus on this value]: if a binding site has nanomolar concentrations in vivo, then the radioligand should ideally have subnanomolar affinity.26 What, then, is the binding potential for P-gp and a high-affinity inhibitor (e.g., tariquidar) in the whole brain? Estimation of Efflux Transporter Density in the Whole Brain (Bmax). To calculate the binding potential, an estimate of P-gp density in the whole brain (Bmax) is required. Although the affinities of radiolabeled inhibitors for P-gp have been experimentally determined, the Bmax has not [reports have been published on P-gp expression in cells of the brain parenchyma expressing P-gp (particularly during neuroinflammation) and in infiltrating lymphocytes]. However, for the purposes of this discussion we will assume that P-gp expression in 2224

dx.doi.org/10.1021/mp400011g | Mol. Pharmaceutics 2013, 10, 2222−2229

Molecular Pharmaceutics

Review

the brain occurs in the brain capillary endothelial cells only. Ordinarily, in vitro studies with brain homogenate are used to determine an approximate Bmax value,9 but the high lipophilicity of the P-gp inhibitors (e.g., clog P tariquidar = 6)27 has made it difficult to estimate this value using this method. Instead, we use the values published recently by Terasaki and co-workers, who measured the expression of transporters in isolated brain capillary endothelial cells using quantitative liquid chromatography-tandem mass spectrometry. This method allows for measurement of protein levels, given as fmol/μg of cellular protein in endothelial cells isolated from brain capillaries of mice,28 monkeys,29 and humans30 (summarized in Table 2). From these studies, we know the concentration of transporters in human brain capillary endothelial cells, but not in the whole brain, which is necessary to calculate Bmax. Therefore, the calculation of the concentration of a protein expressed in capillary endothelial cells in the whole brain is critically reliant on knowledge of the density of brain capillary endothelial cells as a proportion of total brain tissue. This is distinct from the total brain capillary volume, which is much higher, as it includes blood volume. Blood volume in the brain is reported to be approximately 5% of total brain volume,31,32 but the capillary volume is significantly lower. A three-dimensional reconstruction of brain capillaries from frozen serial sections of feline temporoparietal Suprasylvian−Gyrus revealed an average capillary diameter of 4.68 ± 0.92 μm, total capillary length of 42.55 cm/ mm2 of brain tissue, and a capillary volume of 2.8%.33 Yet there is surprisingly little experimental data on the volume occupied by endothelial cells in the brain. The principal values cited come from Pardridge,34 who stated that, “Since the brain capillary endothelial cell volume is only 1 μL/g brain, the endothelial volume in brain is only 0.2% of the total cell volume in brain.” In an earlier report, Pardridge estimated that endothelial cell volume is 0.8 μL per g of brain, given a capillary volume of 1% of total brain volume, and that endothelial cells occupy one-tenth of the total capillary volume (Figure 1).35 Is there experimental support for these approximations? The most useful data appears to come from Hicks and colleagues, who assessed the cross-sectional area of capillary components (basement membrane, endothelial cell, lumen, and pericyte) in the hippocampus and frontal cortex of rats of 3, 9, and 24 months of age.36 Assessment of area was achieved by producing large photographic copies of each electron microscope image, cutting out the components, and weighing the pieces to calculate area. The percentage of area composed of endothelial cells in the frontal cortex was 12% and in the hippocampus was 11% (average basement membrane 4%, lumen 80%, and pericyte 4%). Given an experimental representative brain capillary volume of 2%, and an experimental representative endothelial cell volume that is 10% of the total capillary volume, brain capillary endothelial cell content is indeed estimated at 0.2% of brain volume. Using the numbers from previous studies, we estimate that the Bmax of P-gp in the whole human brain is 1.3 nM (see Appendix A for detailed calculations). The method employed by Terasaki and colleagues to measure the expression of transporter protein did not differentiate between functional transporter at the luminal membrane and internalized transporter; as such, Bmax at the luminal membrane may be slightly lower (as noted earlier, inhibitors may bind both pools of P-gp). We have determined the Bmax values for other brain capillary transporters in a similar fashion (Table 2). It is notable that only two solute carrier transporters demonstrate higher expression in human brain

Figure 1. Diagram of cross-section from a rat brain capillary demonstrating that endothelial cells constitute a small percentage of the total capillary volume. The diagram was generated from electron micrograph published by Hicks and colleagues.36 The scale bar represents 1 μm.

endothelial cells than P-gp. They are associated with energy supply to the brain (glucose transporter 1, Glut1, SLC2A1, calculated brain concentration 29.5 nM) and glutamate supply (glutamate aspartate transporter, Eaat1, SLC1A3, calculated concentration 5.2 nM). While PET functional studies of Glut1 have been reported, and it is expressed only in brain endothelial cells,37 no PET binding studies exist in the literature for this transporter. Binding Potential (Bmax/KD) of P-gp and “High-Affinity” Inhibitors. Using the Bmax value mentioned above for P-gp (1.3 nM) and reported KD values for various P-gp inhibitors, we calculate that Bmax/KD values range from 0.004 to 0.470 (Table 3). These values are well below the Bmax/KD cutoff value of >5− 10 recommended for a useful PET radioligand.26 Although a similar mathematical approach was described in the literature, previous calculations used a vascular volume of 5% rather than a capillary density of 0.2%,15 concluding that the Bmax/KD (i.e., the binding potential) for tariquidar, for instance, is 15, which would Table 3. Calculated Values of Binding Potential (Bmax/KD)a for Various P-gp Inhibitors inhibitor

KDb (nM)

species used for cell assay

reference

calculated Bmax/KD for P-gp

cyclosporin A elacridar tariquidar valspodar zosuquidar

300 2.7 5.1 80 55

hamster human hamster hamster hamster

42 43 14 42 42

0.004 0.474 0.251 0.016 0.023

a

Bmax = concentration of binding sites. KD = equilibrium dissociation constant, where affinity of the ligand for the target is defined as 1/KD. b The KD value for the P-gp inhibitor, laniquidar, has not been reported. 2225

dx.doi.org/10.1021/mp400011g | Mol. Pharmaceutics 2013, 10, 2222−2229

Molecular Pharmaceutics

Review

Figure 2. The high, localized density of P-glycoprotein (P-gp) may transiently reduce the free plasma concentration of [11C]tariquidar in brain capillaries, which may subsequently reduce the concentration of free radioligand in the brain. (A) Uptake of [11C]tariquidar, as measured by positron emission tomography (PET), is low in brains of wild-type and Bcrp knockout, 2-fold higher in P-gp knockout mice, and 9.3-fold higher in P-gp/Bcrp knockout mice. Modified with permission from data published by Langer and colleagues15 (2010). Copyright 2010 Elsevier. (B) Scheme demonstrating the potential interactions of radiolabeled inhibitor [11C]tariquidar (indicated by *) with transporters P-gp (red) and breast cancer resistance protein (Bcrp, yellow). The numbers to the right of each image are representative values for the density (Bmax) of P-gp, free concentration of tariquidar in plasma ([TQR]FP), and free concentration of tariquidar in brain ([TQR]FB) for each mouse strain. The values were calculated assuming that the density of P-gp is 40 000 pM in the capillaries, and that tariquidar has a concentration of 1 nM in the capillary, a KD of 5 nM, and a free fraction of 0.05 (meaning a free concentration of tariquidar of 0.50 pM). In wild-type and Bcrp knockout mice, for every one molecule of tariquidar that is free, eight molecules are bound to P-gp; this binding further decreases the free concentration of tariquidar in plasma to 0.06 pM. In P-gp knockout mice, the free concentration of tariquidar is not further reduced by high-affinity binding to P-gp, although overall parenchymal uptake is still low because the inhibitor is efficiently effluxed as a substrate of Bcrp. In P-gp/Bcrp knockout mice, the radiolabeled molecule is taken up in the brain and trapped (probably by lysosomal trapping7).

High, Localized Transporter Density. Although conventional PET studies of receptor density in the brain require brain penetration of a radioligand,25 PET studies of transporter density do not, as the target (i.e., efflux transporters) is located within capillary walls. At this location, which is exposed to the plasma compartment, efflux transporters can have a direct effect on the free concentration of radioligand. Because only free radioligand can bind to a target, we speculate that this effect on drug disposition is more likely to affect tracer concentrations of drug than pharmacological doses. We propose that the local density of P-gp in the endothelial cells is so high that it transiently binds a high percentage of radiolabeled inhibitor, thereby lowering the free concentration in the capillary such that a negligible amount would enter the brain. If all P-gp is concentrated in the capillary volume, then its concentration (Bmax) in the capillary space is 40 nM. To illustrate the effect of high, localized density on free ligand concentration,

allow for obtaining a sufficiently high in vivo signal with radiolabeled tariquidar (subsequently corrected38). There are clearly limitations to our calculations because they are based on experimentally determined values derived from a range of species: rabbit (brain capillary density), rat (capillary cross sections), human (capillary transporter density and brain protein density), and hamster (inhibitors’ KD for P-gp). Nevertheless, it is clear that the concentration of P-gp in the whole brain is very low (